Mapping old-growth forests using airborne lidar data and satellite images: how do plot size and rarity affect accuracy?
Graphical abstract
This study investigates the mapping of old-growth boreal forests in Finland using airborne lidar and satellite data. It examines how plot size and the rarity of old-growth forests influence the accuracy of detection. Using a Gaussian process classifier on 176 old-growth and 1082 managed forest plots, the study shows that larger plots and data augmentation improve classification performance. The findings support more accurate planning of field-based inventories, aiding conservation and management of rare and fragmented old-growth forests.
The research explores the technical factors affecting the accuracy of old-growth forest mapping in Finnish coniferous forests. It compares plot-level airborne lidar, Sentinel-2 satellite images, and tree-level lidar features, testing 20 m × 20 m versus 60 m × 60 m plots. Larger plots were more effective because they capture spatial patterns of trees and crown structures indicative of naturalness. Data augmentation, generating additional plots within forest boundaries, further increased classification performance, achieving the highest F1-score of 0.74.
The study also quantifies the effect of old-growth forest rarity on mapping accuracy, noting that accuracy declines as forests become rarer due to uneven regional distribution and land-use pressures. These findings emphasize the need to consider both plot size and forest prevalence when planning inventories and modeling forest naturalness. The approach provides a practical framework for supporting conservation prioritization, improving the identification of ecologically valuable old-growth forests, and informing sustainable forest management strategies.
- Implementation
- Monitoring & Projecting
- Landowners & Practitioners
- Planners & Implementers
- Finland